The eternal youth of azobenzene: new photoactive molecular and supramolecular devices

Azobenzene E–Z isomerization

Photoisomerization refers to the transformation of a compound from one isomeric form to another caused by light irradiation [1]. In most instances, the interconvertible forms are E–Z stereoisomers or ring open-closed structures. A prototypical case of photoisomerization is the light-induced E → Z transformation of the –N=N– double bond in azobenzene (Fig. 1). This reaction is extremely fast, efficient, clean and reversible, and the two isomers exhibit quite different structure and properties [2].

Fig. 1

The E and Z isomers of azobenzene and their photo- and thermally induced interconversion.

Azobenzene was first described by E. Mitscherlich in 1834 [3] and its industrial synthesis dates back to 1856. Its E–Z photoisomerization reaction was identified in 1937 by G. S. Hartley [4]. Since then, a huge number of studies involving azobenzene photoisomerization have been performed. To date, a SciFinder search on azobenzene yields more than 20 000 references and the number of azobenzene compounds registered on CAS exceeds 250 000. Indeed, azobenzene has been a playground for both theoretical and experimental chemists for more than a century.

Although the photochemical properties of azobenzene derivatives are affected by the presence of substituents, most azobenzenes behave as schematically shown in Fig. 1: the E isomer is the thermodynamically stable form (albeit exceptions exist [5]) and is converted to the Z isomer by light irradiation. Such a transformation can occur by two basic pathways: the in-plane increase of one Ph–N=N angle (inversion mechanism), or the torsion of the molecule around the N=N axis (rotation mechanism). The determination of the isomerization mechanism in azobenzene and its derivatives has challenged experimentalists and theoreticians for many years and is still a debated issue [6].

Azobenzene photoisomerization is an ideal process to introduce light-driven functionalities in suitably designed molecular and supramolecular systems in order to develop, for example, photoactive drugs [7], devices [8] and materials [9]. As a matter of fact, molecular tweezers based on the azobenzene unit were the first examples of light-driven molecular machines reported in the literature [10] and nowadays have reached a high level of sophistication [11]. The above mentioned properties, together with the versatility of azobenzene in terms of chemical modification and the possibility of tuning its spectroscopic and photochemical behaviour by the choice of the substituents, are the reasons why this photochromic unit is still widely investigated for both fundamental research and technological applications.

Here we illustrate recent investigations undertaken in our laboratories, aimed at using azobenzene isomerization for photo-inducing and -controlling large-amplitude molecular motions, both under thermodynamic and kinetic viewpoints, in multicomponent (supramolecular) species.

Relative light-controlled unidirectional transit of a macrocycle along an axle

Pseudorotaxanes are supramolecular complexes consisting, at the bare minimum, of a thread-like molecule surrounded by a macrocycle. A main element of interest for pseudorotaxanes is that their assembly and disassembly resembles the threading and dethreading of a needle. Since these processes can be controlled by external inputs, they are relevant for designing functional materials [12], as well as for constructing molecular machines based on rotaxanes, catenanes, and related interlocked compounds [13, 14]. An intriguing challenge in this field is the development of a pseudorotaxane motif in which the threading and dethreading movements are made to occur along the same direction [15–17]. The construction of pseudorotaxanes exhibiting stimuli-controlled relative unidirectional threading and dethreading processes is a crucial step for the realization of processive linear motors based on rotaxanes and rotary motors based on catenanes [18]. Moreover, systems of this kind constitute a first step forward towards the construction of an artificial molecular pump.

The directional control of the motion, which is an essential feature of molecular motors, can be achieved by modulating both the thermodynamics and the kinetics of the transition between the mechanical states of the device, that is, by applying ratchet mechanisms to the design of the systems [19], similarly to what happens for biomolecular motors [20]. A few examples of artificial molecular rotary motors and DNA-based linear motors [21] have been described, and fully synthetic linear motor molecules are available [22, 23]. Such systems, however, are based on sophisticated chemical species and/or their operation relies on a complex sequence of chemical reactions. Therefore, the development of concepts and structures for the construction of linear supramolecular motors that are characterized by simple, efficient and reversible operation and can exploit light energy is still an open challenge.

In this section we describe the operation of a simple supramolecular assembly in which a molecular axle passes unidirectionally through the cavity of a molecular ring in response to photochemical and chemical stimulation [24]. It should be recalled, however, that in solution only the movements of the ring and axle components relatively to one another can be considered; therefore, an equally valid view of the device operation is to consider the transit of the macrocyclic ring along the molecular axle. In fact, this convention will be used hereafter to describe the system.

A minimalist approach to the problem is based (Fig. 2) on a non-symmetric axle molecule comprising three different functional units: (i) a passive pseudo-stopper (S), (ii) a central recognition site (A) for the ring, and (iii) a photoswitchable unit (P) at the other end.

Fig. 2

Strategy for the photoinduced unidirectional transit of a molecular ring along a non-symmetric molecular axle. Simplified potential energy curves (free energy versus ring-axle distance) for the states shown, describing the operation of the system in terms of a flashing ratchet mechanism, are also reported. Adapted by permission from Ref. [27].

The strategy at the basis of the operation of the ensemble is shown schematically in Fig. 2. In solution it can be anticipated that, for kinetic reasons, the ring goes to encircle the recognition site A of the axle by passing exclusively from the side of the photoactive gate in its starting α configuration (Fig. 2a). Light irradiation converts the α-P end group into the β form, which should exhibit a higher threading barrier, and may also destabilize the supramolecular complex (Fig. 2b) [25]. The dethreading of a fraction of the axle molecules in the pseudorotaxane population is thus expected, which occurs by slippage of the ring molecules through the side of the axle carrying the S moiety (Fig. 2c). The system is brought back to its initial state by photochemical or thermal conversion of the β-P gate back to the α configuration (Fig. 2d). Overall, the photoinduced directionally controlled transit of the macrocycle along the axle would be obtained according to a flashing energy ratchet mechanism [19].

Two basic requirements are needed for this strategy to work: (i) the energy barriers for the slippage of the ring through the axle end groups should follow the ΔE_b(α – P) < ΔE_b(S) < ΔE_b(β – P) order, and (ii) the ring should form a more stable pseudorotaxane when the axle has the photoswitchable end group in its α configuration compared with the β one. It is also important that the differences in the kinetic and stability constants are sufficiently large, and that the photochemical interconversion of the P gate between its α and β forms is fast, efficient, and reversible.

Suitable candidates for the roles of the molecular ring and of the A units are, respectively, 24-crown-8-type macrocycles and dialkylammonium centers. According to earlier studies [25, 26], azobenzene and methylcyclopentyl moieties could play the role of the photoactive end group P and the pseudo-stopper unit S, respectively. The strategy shown in Fig. 2 can therefore be implemented with the non-symmetric axle E-1⁺ and the macrocyclic ring 2 (Fig. 3).

Fig. 3

Structure formulas and schematic representation of the molecular axle E-1⁺ and the macrocyclic ring 2.

¹H NMR spectroscopic titration experiments showed that in acetonitrile ring 2 encircles E-1⁺ exclusively by passing from its E-azobenzene terminus. Irradiation of E-1⁺ with UV light affords Z-1⁺ almost quantitatively. The increased hindrance of the azobenzene end group upon photoisomerization forces the macrocycle to slip through the cyclopentyl terminus of Z-1⁺. Noticeably, the E→ Z photoisomerization of the azobenzene end group of 1⁺ takes place efficiently also when it is surrounded by 2. Kinetic control of the threading-dethreading side of 1⁺ can thus be achieved by photoadjusting the steric hindrance of its azobenzene end group.

However, as the stability constants of the [E-1⊂2]⁺ and [Z-1⊂2]⁺ pseudorotaxanes in acetonitrile are identical within errors, the dethreading of 2 from Z-1⁺ cannot be caused by the same photochemical stimulus that triggers the azobenzene E→Z isomerization. Hence, potassium ions were added as competitive guests for 2 to promote the disassembly of the complex. Indeed, the addition of two equivalents of KPF₆ at room temperature causes the complete and immediate dethreading of [E-1⊂2]⁺. The addition of the same amount of K⁺ ions to [Z-1⊂2]⁺ also promotes dethreading which, however, takes place on a much slower time scale (t_1/2 = 51 min) [24]. A comparison with the behaviour of model compounds [24] clearly indicates that the chemically induced disassembly of 2 from Z-1⁺ takes place exclusively by slippage of the ring through the cyclopentyl end of the axle.

The photochemically and chemically driven relative unidirectional transit of the ring along the axle can be summarized with the cycle shown in Fig. 4. First of all, 2 passes through the E-azobenzene side of E-1⁺ to form the pseudorotaxane [E-1⊂2]⁺ (Fig. 4a). Irradiation in the near UV region converts quantitatively [E-1⊂2]⁺ into [Z-1⊂2]⁺ (Fig. 4b), characterized by slow assembly-disassembly kinetics. The successive addition of K⁺ ions promotes the dethreading of 2 from Z-1⁺ by the passage of the cyclopentyl moiety through the cavity of the ring (Fig. 4c). Thermal Z→ E back isomerization eventually regenerates E-1⁺ (Fig. 4d), and addition of an excess of 18-crown-6 (18C6) removes K⁺ from 2, thereby affording the re-assembly of [E-1⊂2]⁺ and the full reset of the system.

Fig. 4

Photochemically and chemically controlled relative unidirectional transit of ring 2 along axle 1⁺. Reproduced by permission from Ref. [27].

It is noteworthy that this supramolecular ensemble, if it would be incorporated in a compartmentalized structure (e.g., embedded in the membrane of a vesicle) and subjected to the correct input sequence, could not be used to ‘pump’ the molecular axle and generate a transmembrane chemical potential because the ring component has two identical faces. A strategy similar to the one just described, however, can be applied to supramolecular assemblies based on three-dimensional non-symmetric macrocycles such as cyclodextrins [28] or calixarenes [15], in which face-selective threading can be realized. In particular, calix[6]arene derivatives are interesting candidates for playing as molecular pores or channels because they can be incorporated in the wall of vesicles [29] and their length can approach the thickness of a bilayer membrane [30]. Despite this limitation, the described system is structurally unsophisticated, easy to synthesize, and can be switched conveniently and reversibly: all these features are essential requirements for any real world application.

Very recently, a modified version of the above described system was shown to perform directional threading-dethreading cycles away from equilibrium under steady irradiation, thus behaving as an autonomous light-driven supramolecular motor and providing an example of dissipative self-assembly [31].

A molecular machine exhibiting photoinduced memory effect

A recently reported multicomponent system in which azobenzene photoisomerization plays a crucial role is the [2]rotaxane 3⁴⁺ shown in Fig. 5 [32]. In this species, orthogonal chemical and photochemical stimuli are used to gain full control on the thermodynamics (i.e., the distribution of the rings between the two sites located along the axle) and the kinetics (i.e., the translation rate of the rings between the sites) of molecular shuttling. A system of this kind is interesting not only from the viewpoint of molecular machines but also for that of signal processing and storage. As a matter of fact, controllable molecular shuttles can be considered as bistable mechanical switches at the nanoscale. While the operation of bistable molecular switches is based on classical switching processes between thermodynamically stable states, the development of molecular memories – which rely on a sequential logic behavior [27] – also requires a control of the rates of the mechanical movement between such states.

Fig. 5

Structure formula of the [2]rotaxane E-3⁴⁺ (top) and schematic representation of its chemically and photochemically triggered memory switching cycle (bottom).

The functional units incorporated in 3⁴⁺ (Fig. 5, top) are: (i) a π-electron-deficient ring; (ii) the π-electron donor recognition sites of the dumbbell component, constituted by a tetrathiafulvalene (TTF) unit and a 1,5-dioxynaphthalene (DNP) unit; and (iii) a photoactive 3,5,3′,5′-tetramethylazobenzene (TMeAB) moiety, located in between the TTF and DNP units, which can be reversibly and efficiently switched between its E and Z configurations by photochemical stimuli. Since the TTF unit is more π-electron rich than the DNP one, the macrocycle prefers to encircle the TTF unit rather than the DNP one in the starting co-conformation of 3⁴⁺ (Fig. 5).

This preference is evidenced by the presence of a charge-transfer absorption band peaking at 850 nm. Upon chemical or electrochemical oxidation of the TTF unit to its radical cation (TTF^·+) form, signalled by the appearance of the TTF^·+ absorption features in the 400–650 nm region, the macrocycle shuttles to the DNP recognition site on account of the electrostatic repulsion caused by the TTF^·+ radical cation and the loss of π-donor-acceptor interactions with the ring. Such a process can be monitored by the disappearance of both the band at 850 nm and the sharp absorption features at around 320 nm, typical of the DNP site not surrounded by the macrocyclic ring.

Steady state and time-resolved UV-visible spectroscopic experiments showed that upon quick chemical reduction of the TTF^·+ unit to its neutral state the ring immediately shuttles back to encircle the TTF site if the TMeAB unit is in the E configuration, whereas it remains trapped on the DNP site if the TMeAB unit has been photoisomerized to the Z isomer prior to the TTF^·+→ TTF back reduction (Fig. 5, bottom).

This behavior can be explained considering that the E→ Z isomerization of TMeAB brings about a large geometrical change capable of affecting substantially the free-energy barrier for the shuttling of the macrocycle along the axle component. The Z-azobenzene unit poses a much larger steric hindrance to ring shuttling than does the E-isomer, in analogy with the results described for the previous example.

In summary, the switching cycle of rotaxane 3⁴⁺ (Fig. 5, bottom) consists of the following steps: (i) oxidation of TTF, causing ring shuttling from the TTF^·+ to the DNP site; (ii) UV light irradiation, converting the TMeAB unit from the E to the Z configuration (gate closed); (iii) back reduction, regenerating the neutral TTF unit with the ring still residing on the DNP unit, and (iv) successive photochemical or thermal Z→ E back isomerization, opening the gate and enabling the replacement of the macrocycle onto the TTF primary recognition site. This last step is clearly evidenced by the fact that the first-order rate constant for replacement of the ring onto the regenerated TTF site in the photoisomerized rotaxane, obtained by monitoring the recovery of the charge-transfer absorption band at 850 nm, is in very good agreement with the first-order rate constant corresponding to the thermal Z→ E isomerization of the TMeAB unit, measured by observing the recovery of the absorption band of the E-TMeAB unit at 344 nm (Fig. 6).

Fig. 6

Time-dependent absorption changes (acetonitrile, 295 K), monitored at (a) 344 nm, (E-TMeAB absorption) and (b) 842 nm (macrocycle-TTF charge-transfer absorption), showing the regeneration of E-3⁴⁺ from the metastable state Z-3⁴⁺. The lines represent the data fitting according to a first-order kinetic equation. Adapted by permission from Ref. [32].

In other words, in a “write-lock-erase” experiment based on the cycle shown in Fig. 5 the data is written on the rotaxane by an oxidation stimulus, and locked by UV light irradiation; after the writing session, the oxidized species can be reduced back to the original form without losing the written data for a remarkably longer time compared to thermodynamically controlled molecular switches. Indeed, the data remain stored for a few hours in the dark at room temperature until the thermal opening of the azobenzene gate occurs. Therefore, 3⁴⁺ operates as a bistable memory element under light-triggered kinetic control. It is also important to note that light irradiation not only locks the data previously recorded by oxidation, but also protects the non-oxidized rotaxanes from accidental writing.